Data-center economics continue to change, as traditional enterprise architectures are increasingly being challenged by new alternative solutions that utilize common hardware components together with software to enable scale-out architectures. Until recently the reliability and performance of these solutions were not well known, leaving their adoption to companies who were willing to accept a greater degree of risk for their storage environments.
However, software based architectures are gaining adoption in hyper-scale and other large cloud deployments. Within enterprises, software based systems are gaining adoption; initially deployed for tier two applications, remote and branch offices, but now moving into tier one workloads and large scale-out uses as well. The common aspect for all these deployment models is the fact they leverage commodity server hardware, with enterprise storage features delivered through specialized software. This concept is known by several names including server-based storage, virtual san and software-defined storage (aka SDS).
This architecture has arisen due to the need for reliable storage that is both highly scalable and able to utilize standard hardware components to provide better economics. This architecture has the ability to scale-out to large capacity and performance levels while maintaining lower costs. Additionally, it can maintain balance between CPU, memory, networking and storage subsystems in order to maintain overall system efficiency.
In this paper, we examine the performance of server-based systems together with software-based storage to find their price-performance levels. Testing utilized SDS running on Intel-based servers. The performance was measured utilizing a virtual-server benchmark, with results that may be compared to other published systems. Test results clearly show that system performance and price-performance improved significantly by using solid-state storage media, and in particular NVMe accessed media.
Download the IOmark-VM-HC Certification report here.